Deploying Applications with zc.buildout
log in to bookmark this presentaton
We build and deploy large Python applications for our customers. These applications are built from many component parts including Python packages and non-python applications and tools. Assembling these pieces can be a tedious and error-prone process. The zc.buildout (Buildout) package was developed to automate this process.
As our use of buildout has matured, we've integrated it with system packaging tools and developed techniques for managing software deployemnt and service configuration. I'll provide a detailed description of the techniques and tools used.
- Configuration files
- Deployment pholisophy: keeping software and service configuration separate
- System Packaging
- Making self-contained source releases (tar balls)
- Creating system packages from source releases with RPM
- Service configuration -- system build outs
- Scope: Python processes, cron, log rotation, run scripts, monitoring, etc.
- Place files in "standard" locations.
- Run by root
- Meta recipes
- Goal: provide only essential information.
- Recipes provide basic level of abstraction.
- Buildout provides basic tools to build on these abstractions, but these tools only scale so far.
- Buildout configuration language is *not* a programming language.
- Meta-recipes provide a much more powerful way to build high-level abstractions over basic recipes.
- Next step: Make me a ...
- Model based definition of complete systems spanning multiple hosts with interlocking services
- Automatic management of resources
- Implemented with ssh, system packaging tools (e.g. yum) and buildout-based tools.
- Four levels of abstraction
- Current status